Scientific Limits — what this lab does NOT establish

Written in the repository's falsification spirit: state the fences before anyone has to find them. Every item here is not-yet-shown, outside-model-scope, requires-experiment, or metaphor-preserved — never proven.

Hard boundaries (asserted nowhere in this lab)

• Ozone is not alive and is not life itself.
• Ozone alone does not prove life (it is produced abiotically — Venus, Mars).
• Ozone alone does not create or retain water (habitability is multivariate).
• Atmospheric pressure does not replace gravity. The pressure-weight model g = k·P, calibrated on Earth, is shown by the test suite to fail on every other body.
• Active inference does not prove any physical mechanism. It frames model comparison.
• Protons are not shown to be a soul. That reading is preserved only as metaphor.
• Humans are not asserted to be non-terrestrial. No such claim is in scope.
• There is no unlimited energy. Vacuum is an incremental engineering gain; the radiative loss term and the Shockley–Queisser limit are not erased.
• A simulation does not prove reality.

Model-fidelity limits of what IS implemented

• SP.Lab.Radiation is a single-absorber Beer–Lambert toy: no scattering, no multi-species absorption, no solar-zenith-angle or altitude structure. It shows the direction and rough magnitude of ozone shielding, not a radiative-transfer solution.
• SP.Lab.PlanetaryData values are agency measurements (evidence class B), read as data. Surface gravity carries small rotation/oblateness effects the GM/R² model omits (this is why the test tolerance is 2%, not exact).
• SP.Lab.Bioenergetics viability thresholds (water 0.6, PMF 50 mV, radiation 1.0) are engineering choices (evidence class C), declared in thresholds/0. They set the toy cell's behaviour; they are not measured biological constants.
• SP.Lab.ModelCompare weights are a declared rubric (class C), a bookkeeping aid. The load-bearing quantity is the measured failure count, not the rubric score.
• SP.Lab.SolarEnergy is a steady-state lumped heat balance: no transient dynamics, no spectral detail, no real device geometry.

Specified but NOT yet implemented (owed, not hidden)

• Full ozone Chapman kinetics (rate equations) — only the shield layer is coded.
• Diffusion-limited hydrogen escape (Hunten flux) — escape_velocity is coded; the full flux is documented in lab/proofs/water_escape.md and not yet a tested module.
• Generational DNA/inheritance lab layer — specified; the host repo's SP.Genome exists, a lab-specific inheritance experiment does not.
• The UI/UX experience — lab/ui/ is a spec + static stub, not a finished interface.
• A live active-inference agent loop in the lab — the host system implements active inference (SP.*); the lab only documents the bound discipline.

The one most important open question

Whether any of these toy models, assembled, can make a novel, pre-registered, out-of-sample prediction that a mainstream baseline cannot — rather than re-describing known values. Until such a prediction is registered before its test and survives, the lab's status is: a faithful, falsifiable re-derivation of established science, in which one fringe thesis (pressure-gravity) is contradicted and one (ozone-as-shield) survives only in its narrowed form.